monk1337

import tensorflow as tf
import numpy as np

def attention(inputs, attention_size, time_major=False, return_alphas=False):

    if isinstance(inputs, tuple):
        inputs = tf.concat(inputs, 2)

    if time_major:
        inputs = tf.array_ops.transpose(inputs, [1, 0, 2])

monk1337

def contrastive_loss(self, y, d, batch_size):
        tmp = y * tf.square(d)
        # tmp= tf.mul(y,tf.square(d))
        tmp2 = (1 - y) * tf.square(tf.maximum((1 - d), 0))
        return tf.reduce_sum(tmp + tmp2) / batch_size / 2

    
       #  self.scores = logits

        with tf.name_scope("loss"):

monk1337

5                                              # The shape is []
[ 1., 2., 3., 4. ]                             # The shape is [4]
[[ 1., 2., 3., 4. ],[ 5., 6., 7., 8. ]]        # Matrix of shape [ 2,4]
[[[ 1., 2., 3., 4. ] ],[ [ 5., 6., 7., 8. ]]]  # Tensor of shape [ 2, 1, 4 ]

monk1337

import torch
print('one dim')
print(torch.rand(1))

# output: 

# one dim
# tensor([ 0.3725])

monk1337

import torch
print('two dim')
print(torch.rand(2,5))

# output:
# two dim
# tensor([[ 0.2495,  0.2948,  0.5486,  0.5077,  0.1657],
#         [ 0.4142,  0.8142,  0.8635,  0.8827,  0.7176]])

monk1337

#normal distribution with mean=0 var=1
print(torch.rand(1,20,dtype=torch.float))
print(torch.rand(1,20,dtype=torch.double))

# output:
# tensor([[ 0.4976,  0.5590,  0.4242,  0.3130,  0.4160,  0.2188,  0.5643,
#           0.8620,  0.6020,  0.0883,  0.2870,  0.5136,  0.8119,  0.7638,
#           0.3188,  0.1961,  0.3527,  0.9613,  0.2914,  0.3882]])
# tensor([[ 0.9707,  0.9771,  0.0904,  0.0374,  0.7983,  0.4952,  0.2216,

monk1337

#best part with pytorch is you can treat pytorch object as python object 
var=torch.rand(5,2,dtype=torch.double)

#we can loop over it
for i in var:
    for k in i:
        print(k)
        
# output:
# tensor(0.9142, dtype=torch.float64)

monk1337

tensor=torch.Tensor(4,3)
print(tensor)

# output:
# tensor([[ 0.0000e+00,  1.0842e-19,  6.0390e+35],
#         [ 2.8586e-42,  4.2039e-45,  0.0000e+00],
#         [ 0.0000e+00,  0.0000e+00,  0.0000e+00],
#         [ 0.0000e+00,  0.0000e+00,  2.7551e-40]])

monk1337

print(tensor.size())

# output:
# torch.Size([4, 3])

monk1337

torch.cuda.is_available()

#output
# False

#use of cuda
print(torch.Tensor(1,2).cuda())

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)